News coverage and the social dimension of belief reinforcement: a longitudinal mixed-methods approach

Abstract

Understanding the conditions behind media-driven belief maintenance and reinforcement is critical for a comprehensive account of long-term media effects. Focusing on news coverage and beliefs about crime developments in Sweden, this study addresses the so-called “filtering function” of interpersonal communication: the idea that media messages and beliefs are validated in social networks. Using a longitudinal mixed-methods approach—combining content analysis of news coverage, a six-wave panel survey, and (focus) group discussions—the study analyses the long-term processes as well as the social validation mechanisms embedded within interpersonal discussion networks to understand belief reinforcement over time. Both the quantitative and the qualitative data support the basic social validation mechanism underpinning reinforcement effects, suggesting several distinct ways in which news coverage and beliefs are validated (and rejected) in social communication. These findings contribute to research on dynamic media effects, cultivation theory, and social networks.

Understanding the conditions behind media-driven belief maintenance and reinforcement is critical for a comprehensive account of long-term media effects. Several recent theoretical developments highlight the need to address the dynamic and conditional nature of communication effects—beyond the very short-term influences that are typically studied (Valkenburg & Peter, 2013; Slater, 2015; Shehata et al., 2021; Thomas, 2022). How media coverage and use contribute to long-term belief maintenance and reinforcement has been a key issue in cultivation research since the theory was originally formulated (Gerbner et al., 1986; Potter, 2014)—but the topic has also received growing attention lately (Long, 2023; Slater, 2015; Shehata et al., 2024).

Despite growing attention, however, research has not yet comprehensively studied these influences empirically. This is due to conceptual ambiguities as well as limited research designs and analytical approaches. Focusing on broad cultivation-like influences, the key theoretical argument addressed in the current study is that long-term reinforcement (and maintenance) effects depend on how people talk about issues in their social networks. This basic idea is not new. Research on two-step flows (Brosius & Weimann, 1996, Katz & Lazarsfeld, 1955) and, more specifically, the “filtering hypothesis” (Neiheisel & Niebler, 2015; Schmitt-Beck, 2003) has particularly emphasized the critical role of social networks and interpersonal communication in disseminating and validating mass media messages. More recently, this filtering function of interpersonal discussion has also been linked to long-term effect theories such as the Reinforcing Spirals Model (RSM) (Slater, 2015; Song & Boomgaarden, 2017). Empirical research is, however, very scarce. Both with respect to the long-term processes of these dynamics and the social validation mechanisms embedded within interpersonal discussion networks theorized to condition belief reinforcement over time.

To capture these effect dynamics, we thus need to simultaneously consider both the short-term social mechanisms of influence as well as the long-term belief developments. Following the classic case of cultivation research, this study does so by focusing on news coverage and beliefs about crime developments in Sweden, during a time when this issue gained significant public attention. Combining data from (1) a media content analysis, (2) a six-wave panel survey, and (3) group discussions, conducted in Sweden over a period of almost three years (2020–2022), the study analyzes both the long-term processes and social validation mechanisms of interpersonal conversations relevant to media-driven belief maintenance and reinforcement. As such, while the study focuses on belief outcomes and longitudinal effect processes primarily described by cultivation theory, the study integrates arguments from related theories to address the role of social communication as a form of resonance mechanism—either validating or rejecting dominant media messages.

The next section outlines the main theoretical argument linking recent conceptualizations of dynamic media effects to the filtering hypothesis and theories of social communication. Apart from conceptualizing belief reinforcement and maintenance as media effects and discussing the conditions under which they are more likely, specific attention is given to interpersonal communication as a social context for belief validation. The case studied—news coverage and beliefs about crime—is then discussed in greater detail along with a set of hypotheses and research questions. After presenting the empirical study, the concluding section elaborates on the theoretical and societal implications of the main findings.

Longitudinal media effect dynamics

Media effects research increasingly emphasizes the importance of theorizing and testing more complex longitudinal effect dynamics—going beyond studies of short-term attitude changes (Baden & Lecheler, 2012; Geiß, 2022; Shehata et al., 2021). The growing interest in understanding belief stability as an outcome of communication reflects the revival of some classic concepts in the field of media effects research (Bennett & Iyengar, 2008; Holbert et al., 2010; Klapper, 1960; Long, 2023)—including cultivation theory (Cöster & Shehata, 2024; Gerbner et al., 1986). Only recently, however, has the field moved in a direction of thoroughly theorizing more complex effect dynamics. For instance, theoretical developments such as the Differential Susceptibility to Media Effects Model and the RSM emphasize the importance of understanding communication effects as both (1) long-term, dynamic, and transactional, on the one hand, as well as (2) conditional and interactive, on the other hand (Slater, 2015; Valkenburg & Peter, 2013). Importantly, the current study does not specifically address these two theories but builds upon them to conceptualize and better understand long-term belief reinforcement. Below I first discuss belief maintenance and reinforcement as two distinct long-term effect dynamics seldom captured in empirical research. Next, the role of interpersonal communication as a social validation mechanism is elaborated considering these two effect dynamics.

Belief maintenance and reinforcement as media effects

Belief maintenance and reinforcement reflect two long-term dynamics typically thought of as belief stability (Long, 2023; Shehata et al., 2021; Slater, 2015). These effect dynamics are, however, seldom convincingly captured in empirical research—due to both conceptual ambiguities and analytical limitations. Theoretically, they both build on a similar mechanism—belief validation—according to which an already stored belief is validated through exposure to belief-congruent communication, leading to confirmation of the pre-existing belief (Potter, 2011; Shehata et al., 2021; Slater, 2015). They differ however in terms of their respective long-term outcomes: belief stability (maintenance) versus belief intensification (reinforcement).

On the one hand, belief reinforcement is defined here as (1) a long-term process of (2) intra-individual belief intensification over time. The “long-term” specification conceptualizes reinforcement as something qualitatively different from what most empirical studies actually capture: temporary time-specific deviations in belief positions. The “belief intensification” specification clarifies that reinforcement should be understood as a process by which a pre-existing belief position becomes more extreme over time. Belief maintenance, on the other hand, reflects (1) a long-term process of (2) intra-individual belief stability over time. Rather than positional movements toward the endpoints of a belief scale, maintenance entails holding on to the same belief position. As both belief maintenance and reinforcement represent general dynamics potentially caused by multiple factors, they must also be (3) driven by communication to be considered effects of communication. Thus, to count as a communication effect, communication must make a causal difference. Recent theoretical and methodological work has developed approaches to empirically distinguish and capture these dynamics, given appropriate research designs (Shehata et al., 2023; Thomas, 2022; Thomas et al., 2021).

Belief validation is thus the key mechanism behind these effect dynamics. Individuals hold a belief, which is validated through communicative practices, leading to belief maintenance or reinforcement over time. Theory and research suggest that belief maintenance and reinforcement are most likely on established issues for which people already possess pre-existing beliefs (Druckman & Leeper, 2012; Zucker, 1978). These effects could be driven either by motivated selective exposure, when individuals seek-out and engage with belief-congruent media messages (Knobloch-Westerwick, 2012; Slater, 2015), or by intense and consonant one-sided news environments, where most major media report issues or events in similar and consistent ways—a form of long-term effect process described by cultivation theory (Gerbner et al., 1986; Morgan et al., 2012; Shehata et al., 2022). At the same time, belief maintenance and reinforcement are generic outcomes not specifically tied to a particular theory of media effects. However, belief validation also has a strong social component. How citizens talk about societal problems and news they encounter in the media is likely to have a significant impact on message acceptance and issue interpretation. Both classic concepts in mass communication research (Katz & Lazarsfeld, 1955; Neiheisel & Niebler, 2015; Schmitt-Beck, 2003) and more recent theories of conditional media effects (Slater, 2015; Song & Boomgaarden, 2017) emphasize the filtering function of social networks as people make sense of news coverage, media messages and the “world outside”.

Social belief validation and the filtering hypothesis

While the news media can have a substantial impact on citizens’ perceptions of societal conditions and problems (Lecheler & de Vreese, 2019; McCombs & Valenzuela, 2020), social networks and interpersonal communication may condition such influences. Extensive news coverage of established societal issues may lead to belief maintenance or reinforcement over time, depending upon how those issues are talked about and discussed in everyday social settings (Schmitt-Beck & Lup, 2013; Song & Boomgaarden, 2017; Southwell & Yzer, 2007).

According to the classic two-step flow hypothesis, information from the news media is picked up by more engaged segments of the population, who subsequently pass these messages on by talking to others in their social networks (Brosius & Weimann, 1996; Katz & Lazarsfeld, 1955). However, interpersonal communication is not only about pure transmission of messages but entails a broader social function as well. This is reflected by Southwell and Yzer (2007) who regard “conversations not just as simple information delivery between people but rather as relationally and socially consequential behavior” (p. 423). This is also the core idea behind the “filtering hypothesis”, originating from the work of Katz and Lazarsfeld (1955). As summarized by Schmitt-Beck (2003) political conversations “fulfils a crucial ‘meta-communicative’ function by telling them whether or not media messages are valid, and whether or not they should therefore be accepted” (p. 235; see also Conover & Miller, 2018; Schmitt-Beck & Lup, 2013). Through conversations with others, citizens explore the viability and validity of media messages and beliefs. If validated by people in their surroundings, they are more likely to accept these messages. If not validated, but rather questioned or criticized, they are more likely to reject messages and reconsider beliefs. This is thus a social and inter-personal validation process—which has been directly associated to processes of belief or attitude reinforcement (Katz & Lazarsfeld, 1955; Neiheisel & Niebler, 2015; Song & Boomgaarden, 2017).

The basic idea of the filtering hypothesis is thus straightforward: talking about societal issues in ways that align (resonate) with patterns of news coverage should promote message acceptance and—by extension—belief maintenance or reinforcement. Exactly how such mechanisms of message validation and rejection operate in social communication is less clear, however. The more precise conversational and social-interactive practices often remain hidden in quantitative research. Opening the “black box” of these social dynamics of message and belief validation requires in-depth analysis of actual conversations among citizens (Conover & Miller, 2018; Gamson, 1992; Glynn et al., 2016). Previous work on message reception, elaboration, and political talk, suggests three dimensions as relevant for understanding how news coverage is socially validated or rejected during conversations.

The first dimension refers to how conversations relate to the news media as an (authoritative) source of information. This is reflected in how citizens talk about reality, societal conditions, and developments—rather than what they explicitly say about their trust in the news media (Gamson, 1992; Neuman et al., 1992). Liebes and Katz (1986) distinguish between referential and critical orientations toward the media, reflecting the extent to which audiences perceive the content they consume as having real-life relevance. Thus, by “uncritically” referring to, and making use of, mediated reality as a true and accurate account of societal conditions during conversations, interpersonal communication both passes on and validates the narratives promoted by the media. Conversely, citizens may refer to media coverage in more critical ways, casting doubt over or questioning the validity of the news medias’ representations of reality.

The second dimension refers to how conversations relate to other sources of information. Although research suggests that people rely heavily on the news media when few alternative sources are available (McCombs & Valenzuela, 2020; Shehata et al., 2021), qualitative studies demonstrate that citizens draw from and integrate various forms of experiences and knowledge when talking about societal matters (Gamson, 1992; Glynn et al., 2016). In a broad sense, people may talk about societal developments in ways that imply consistency between news coverage and other sources of information (resonance), or in ways that contrast and suggest inconsistency (dissonance). While the former functions as cross-validation reinforcing media messages, the latter creates an impression of multiple alternative representations of reality. In both cases, however, new (external) information is added and integrated into the conversations. While cultivation theory emphasizes the importance of resonance between media messages and citizens’ personal experiences as a form of validation mechanism (Morgan et al., 2012; Shrum, 2017), resonance can also refer to broader sets of knowledge that people bring to conversations (Gamson, 1992; Shehata et al., 2021).

The third dimension refers to how agreement and disagreement is socially communicated and negotiated. Here focus shifts from the content of what is being said, to discussants’ reactions to those statements—as social manifestations of message validation and rejection. While people tend to talk about societal matters primarily in homogenous close-tie settings of likeminded people, conversations are not dichotomous or static (Ekström, 2016; Mutz, 2006). Rather they are fluid and dynamic, moving back-and-forth between areas of potential agreement and disagreement (Conover & Searing, 2005). Studies also suggest that people are more willing to express and explore disagreement in safe close-tie settings (Morey, Eveland & Hutchens, 2012). Accordingly, in-depth analysis of political discussions shows that “public opinion can be exceptionally dynamic and complex, that opinions are constructed and reconstructed in particular social contexts, and that opinions emerge from a synthesis of personal, social, and mass media information” (Glynn et al. 2016, p. 72; Wilson & Hodges, 1992). Thus, the ways in which expressed beliefs are socially validated or rejected in manifestations of agreement-disagreement is potentially critical for message acceptance.

The case: negative news and crime perceptions in Sweden

This study addresses the long-term processes as well as the social validation mechanisms embedded in interpersonal discussion networks to understand media-driven belief maintenance and reinforcement. It does so by focusing on a likely case for such media effect dynamics: news coverage of and public beliefs about crime developments in society. There are theoretical as well as empirical reasons for this. In terms of theory, crime is a classic in the media effects literature, of relevance not only to cultivation theory, but to agenda-setting and framing as well (Lecheler & de Vreese, 2019; McCombs & Valenzuela, 2020; Shrum, 2017). As a topic, crime is a well-established societal problem, salient in the media, but still an issue where news coverage is expected to matter for public opinion. The more the news media focus on crime and frame societal developments in a negative way, the more individuals are likely to see this as a societal problem (Mutz, 1998). At the same time, crime is not unfamiliar to citizens. Most people already hold pre-existing beliefs and well-developed schemas related to crime, which should matter for short- and long-term belief dynamics (Baden & Lecheler, 2012; Shehata et al., 2021).

Furthermore, the specific Swedish empirical context points towards certain belief dynamics as being more likely. Violent crimes have become an increasingly salient societal problem in Sweden in the last decade—in the media, among politicians and the public. Public opinion data from the annual Swedish SOM surveys show, for instance, that “law and order” has gained attention as the most important problem (MIP) since 2014–2015. In 2014, 4% mentioned this as the MIP, followed by a gradual increase until 2019 when 28% of Swedish citizens named this issue. Since then, law and order has been in top of the MIPs, with 34% naming it in 2020, 41% in 2021 and 37% in 2022 (Swedish Trends, 1986–2022). During the same period, actual crimes rates are somewhat more complex. With respect to instances of lethal violence, there was a slight decline in the total number of confirmed cases, from 124 in the year 2020, to 113 in 2021, and 116 in 2022. On average, however, the number of confirmed cases of lethal violence was higher during the five-year period between 2018 and 2022 (114 cases a year) than during the previous period, 2013–2017 (101 cases a year). Thus, there is long-term trend of a growing number of confirmed cases, but a slight decline during the years of the current study. Furthermore, while the total number dropped, there is a clear trend toward increasing use of firearms, which was used in 48 instances in 2020, up to 63 cases in 2022—reflecting a change in the nature of violent crimes (Brå, 2023a). A somewhat similar trend is evident when looking at the total number of reported offences, which increased up until 2020, and declined in 2021 and 2022—a pattern reflected also with respect to reported “crimes against persons” (Brå, 2023b). One of the institutes summarized current trends as reflecting fewer but more severe types of crimes (Kunskapsverket, 2023).

Overall, these conditions suggest a likely case for people to express (increasingly) pessimistic beliefs with respect to crime developments in Sweden between early 2020 and late 2022 (i.e., perceiving crime developments in negative terms, as a major societal problem, etc.). Whether the process reflects belief maintenance or reinforcement over time is unclear, however, and should partly depend on how the news media report on crime during these times. In general, crime coverage is likely to be dominated by a negative tone, due both to the problem-oriented character of the issue itself and to the inherent negativity biases of news reporting (Cöster & Shehata, 2024; Lengauer et al., 2012). All else equal, however, belief maintenance is most likely in stable news environments, while belief reinforcement is more likely when news coverage intensifies in terms of salience and/or negative tone (Slater, 2015; Shehata et al., 2023; Wilson & Hodges, 1992).

RQ1: What is the overall tone (negative/positive) of news coverage of crime between 2020 and 2022?

RQ2: How are trends in news coverage related to trends in crime beliefs (CB) and interpersonal crime communication (IPC) over time?

According to the filtering hypothesis, news coverage is most likely to influence beliefs when people talk about crime in ways that resonate with media messages (Neiheisel & Niebler, 2015; Schmitt-Beck, 2003). This study addresses this argument in two ways, focusing both on the long-term processes of these effect dynamics and on the social validation mechanisms at work during conversations. Importantly, the belief dynamic of interest here concerns the role of interpersonal communication in relation to (changes in) broad patterns of news coverage across major media outlets, rather than belief reinforcement processes driven by motivated selective exposure to belief-congruent news outlets. As such, the type of reinforcement (and maintenance) addressed here is more similar to cultivation-like processes than to the reciprocal dynamics described by the RSM (Gerbner et al., 1986; Morgan et al, 2012). With respect to the long-term processes, engaging in interpersonal communication aligning with the dominant tone of news is expected to reinforce pre-existing beliefs—in the sense that conversations about crime focusing on negative (positive) societal developments, should lead to belief intensification (reinforcement) of negative (positive) crime beliefs over time. Whether these effects lead to a long-term process of belief maintenance or reinforcement is however less clear, since such dynamics depend on the development of news coverage over time, as noted above.

H1: Interpersonal communication (IPC) has immediate, short-term, effects on crime beliefs (CB), in the sense that talking about crime in negative terms increases pessimistic beliefs about crime developments in society, and vice versa.

RQ2: Does the dynamic relationship between interpersonal communication (IPC) and crime beliefs (CB) reflect a long-term process of belief maintenance or reinforcement?

With respect to the social validation mechanisms behind the filtering hypothesis, theorized to support the long-term maintenance and reinforcement effects of interest here, the study digs deeper into the ways in which crime news and beliefs are validated during conversations.

RQ3: How are crime beliefs (CB) validated during conversations, with respect to how discussants (a) relate to news coverage about crime, (b) integrate other sources of information about crime, as well as (c) socially express and negotiate agreement and disagreement?

In sum, the argument addressed here reflects a process of belief reinforcement (or maintenance) involving both short- and long-term dynamics. First, broad patterns of crime coverage in major media outlets are expected to influence how citizens’ talk about crime in their everyday lives (Brosius & Weimann, 1996; Djerf-Pierre et al., 2024). Second, through various filtering mechanisms within such social communication, messages and beliefs are validated, potentially contributing to reinforcement of crime beliefs. These effects are assumed to operate instantaneously (short-term), following accessibility-based cognitive theories of communication effects (Baden & Lecheler, 2012; Shrum, 2017; Slater, 2015), but also build up cumulatively over time contributing to belief reinforcement (long-term). The study aims at addressing both these dimensions of belief reinforcement.

Data and methods

To address the processes and social-conversational practices of belief maintenance and reinforcement, this study combines three sources of data: (a) a media content analysis of crime news matched with (b) a six-wave panel survey, as well as (c) a group discussion study. The panel survey and group discussion studies were designed specifically to enable comparisons while addressing qualitatively different aspects of belief reinforcement. While the content analysis and panel survey primarily address the long-term belief dynamics, the group discussion study seeks to uncover the social mechanisms of message and belief validation as people talk about crime and crime developments in society.

Analyzing the process of belief reinforcement

To capture the long-term processes of belief reinforcement and maintenance, the quantitative part of the study combines an extensive content analysis of crime news reporting in six major Swedish news media outlets with a six-wave panel survey addressing crime beliefs, interpersonal communication and news habits among Swedish citizens. The longitudinal design aimed at striking a balance between being able to capture long-term developments over several years, on the one hand, and identify more short-term (immediate) influences, on the other hand. The theoretical basis for the design follows cognitive (memory-based) accounts of cultivation and reinforcement effects suggesting that while influences operate immediately (short-term), by increasing temporary accessibility, effects may also build up cumulatively over time and, thereby, have more long-term consequences (Morgan et al., 2012; Shrum, 2017; Slater, 2015). To enable this, we use a longitudinal design covering a period of almost three years (from early 2020 to late 2022), with time-lags between panel waves of approximately 6 months, and survey items designed to capture both short- and long-term influences (see communication measures below).

The panel survey was conducted in collaboration with the Laboratory of Opinion Research (LORE) at the University of Gothenburg. LORE hosts a standing panel of probability-recruited web survey participants initially contacted through regular mail on an annual basis. A sample of 3,327 probability-recruited panelists—pre-stratified on gender, age, and education—was invited to Wave 1 of the panel (March–April, 2020)—out of which 2,058 panelists responded (Gross Completion Rate = 61.9%; Net Completion Rate = 64.1%). The invited sample was then invited to five additional waves of data collection, with time-lags of approximately six months in between. Wave 2 was conducted in October–November 2020 (N = 1,700; GCOMR = 54,2%; NCOMR = 56.7%), Wave 3 in April–May 2021 (N = 1,551; GCOMR = 51,5%; NCOMR = 54.5%), Wave 4 in October-November 2021 (N = 1,491; GCOMR = 50,4%; NCOMR = 53.4%), Wave 5 in April-May 2022 (N = 1,341; GCOMR = 46,8%; NCOMR = 50.1%), and Wave 6 in October-November (N = 1,297; GCOMR = 46,2%; NCOMR = 49.8%). Supplementary Appendix Table A5 presents a closer analysis of panel dropout, showing that women were less likely to participate in all six panel waves, while older, highly educated and politically interested respondents are more likely to participate in all waves.

The survey measured three main constructs. Following cultivation research, crime beliefs were measured as specific perceptions about country-level conditions and developments—ie, so-called “sociotropic perceptions” (Djerf-Pierre et al., 2024; Mutz, 1998) – using a battery of three items, following the survey question “Various claims are sometimes heard in public discussions about violent crimes. To what extent do you agree with the following?”. The statements included were (a) “Violent crimes have decreased in Sweden during the last years”; (b) “More violent crimes are committed per person in Sweden than in most other countries in the EU”; and (c) “Sweden currently has very large problems with violent crimes”, with response categories ranging from 1 (“Not true at all”) to 7 (“Completely true”). Item 1 was reversed before constructing an averaged index of these three items, with higher values representing more negative perceptions of crime developments (Cronbach’s α: W1 = 0.77, W2 = 0.78; W3 = 0.76; W4 = 0.78; W5 = 0.78; W6 = 0.77). Interpersonal crime communication was measured by combining two items tapping the (a) frequency and (b) tone of everyday crime talk. Respondents were first asked how often they usually talk about crime with family or friends, with a response scale ranging from 1 (“Every day”), 2 (“Several times a week”), 3 (“At least once a week”), 4 (“At least once a month”), 5 (“More seldom”), to 6 (“Never”) – thereby sensitive to potential changes in frequency of talk between panel waves. A follow-up question asked about what they usually talk about when discussing crime with family or friends, with a response scale ranging from 1 (“Mostly things that work badly”), through 4 (“Both good and bad things”) to 7 (“Mostly things that work well”). An index was constructed by first reversing both items and then multiplying frequency with tone, yielding an interpersonal talk measure reflecting the extent to which respondents talk frequently about negative developments. The strength of this measure is that it captures not only how often people talk but the character or content of these discussions. Finally, the survey measured exposure to a range of established news outlets in Sweden, including public service radio (Ekot, SR), public service television (Rapport, SVT), commercial television (TV4 Nyheterna), two tabloid newspapers (Aftonbladet and Expressen) as well as two broadsheets (Dagens Nyheter and Svenska Dagbladet). Exposure to these in both traditional and online forms were measured as the frequency of use in the past month, ranging from 1 (Daily) to 6 (Never). The retrospective timeframe of these news use measures reflects the timeframe of the content analysis.

To match the panel survey, crime news reporting in the six media outlets were content analyzed using manual coding. News items published in the selected outlets every other day one month prior to the launch of each panel wave were included (14 days for each wave). For newspaper articles, the media archive Retriever was used to identity relevant items based on an elaborated search string focusing on crime-related news. For broadcast media (Ekot, Rapport and TV4 Nyheterna), the selected news programs were screened manually to identify all stories with crime as the main topic. A total of 1,681 stories were manually coded, with the number of news items suggesting a general increase in salience over time (W1 N = 193; W2 N = 257; W3 N = 293; W4 N = 399; W5 N = 236; W6 N = 303). For this study, the tone of news reports was coded to capture the extent to which the media provided a negative or positive account of crime developments in Sweden. Three variables were coded: (a) the presence of a negative tone (Yes/No), (b) the presence of a positive tone (Yes/No), and (c) the balance between negative and positive tone (Negative/Neutral/Positive). Coding was conducted by a trained coder and inter-coder reliability assessed at the outset using a second coder recoding a subsample of 100 news items. The test showed decent reliability scores across the three variables (Krippendorff’s α: Negative tone = 0.83; Positive tone = 0.81; Balance = 0.79).

To empirically address the long-term process of belief maintenance and reinforcement as outlined in the theory section, we need a statistical model that captures the main properties of interest. More specifically, an approach that models (a) long-term belief trajectories, (b) effects of interpersonal communication on crime beliefs, and (c) separates within-person from between-person dynamics. Recent work on statistical modeling for reciprocal communication effects suggests a combination of the Random-intercept Cross-lagged Panel Model (RI-CLPM) and the Latent Curve Model with Structural Residuals (LCM-SR) as most appropriate in terms of meeting the relevant theoretical criteria (Curran et al., 2014; Hamaker et al., 2015; Shehata et al., 2023; Thomas et al., 2021). The main difference between the RI-CLPM and the LCM-SR is that the latter explicitly models a latent random slope factor (S_yi) in addition to the latent random intercept (I_yi) and thereby provides an estimate for the long-term development of a factor. The LCM-SR is therefore the most appropriate model here.

The LCM-SR used here is illustrated in Figure 1, focusing on the two variables of main interest in this study: interpersonal communication and crime beliefs. The between-person (inter-individual) part of the model is represented by the latent intercepts and slopes for beliefs (I_yi and S_yi) and talk (I_xi and S_xi). These capture between-person differences in levels and long-term developments over time for both factors. This part reveals whether there is any evidence for long-term reinforcement or maintenance of crime beliefs over time, as well as whether belief trajectories are related to developments in talk. The within-person (intra-individual) part of the model estimates the autoregressive and reciprocal effects between time-specific deviations from individual baselines of beliefs (e_y) and talk (e_x). This allows for stronger causal claims by providing pure within-person evidence of effect dynamics over time. In contrast to the standard LCM-SR, however, I estimate instantaneous, rather than lagged, effects of talk on beliefs. This modification follows from the combination of theoretical as well as methodological considerations. Theoretically, cognitive accounts of communication effects suggest that influences operate relatively immediately following exposure, subsequently fading or strengthening depending on the nature of additional exposure (Chong & Druckman, 2007; Lecheler & de Vreese, 2016; Shehata et al., 2024). For instance, cultivation effects have been conceptualized as long-term in nature, but they also operate instantaneously by increasing accessibility of beliefs (Shrum, 2017). Accordingly, modelling contemporaneous communication effects thus provides a better theoretical match than lagged effects, using a survey design with time-lags of approximately 6 months.

Analyzing the social mechanisms of belief validation

To address the social mechanisms of interpersonal communication about crime, (focus) group discussions were conducted in the beginning of 2020. Focus groups are especially well-suited for studying social interactive mechanisms in communication, by allowing “researchers to observe social dynamics that are invisible in many other methods” (Gamson, 1992; Glynn et al., 2016, p. 72). In particular, since “individual participants can influence one another, focus groups can also echo the social context within which people discuss public affairs” (Conover et al., 2005, p. 273). In terms of timing, the group discussions took place a few weeks prior to the first wave of the panel survey. The primary aim of these group discussions was to study how citizens talk and interact during conversations about crime and crime developments in times of growing attention to these matters in public discourse.

To resemble authentic every-day talk as closely as possible, the group discussions were composed of people who already knew each other. The design therefore aimed at within-group homogeneity (friends and acquaintances) and between-group heterogeneity (in relation to broad socio-demographic background variables). The conversations are thereby more likely to mimic real-world dynamics and characteristics than if participants were unfamiliar with each other. Between-group heterogeneity was sought to maximize possibilities for variation in ways of talking about crime. We took specific care to cover groups with different age and educational/professional training—as well as diversity in terms of gender. Strategic sampling of six “node persons” with different background was used. These nodes were then asked to suggest other participants from their own personal network, aiming to create a total of six groups of approximately 4–6 people each. All groups were recruited within the greater metropolitan area of Gothenburg. In total, 37 participants were included (21 men and 16 women) with an age span of 25–80 years old.

Each discussion covered the topic of crime in addition to two other topics (not included in this paper). The group discussions took place at various locations in the Gothenburg region and lasted for 90–120 minutes. The conversations were semi-structured and led by an experienced moderator, who used a guideline with open questions to initiate and thematically organize the discussions. The main themes guiding the talks focused on participants’ “perceptions of crime and crime developments in society”, “how they get informed, hear and learn about crime-related issues”, and “perceptions and evaluations of media reporting of crime”.

Transcriptions of the conversations were inductively analyzed with particular focus on mechanisms of belief validation (and rejection) along the three theoretically derived dimensions of social communication: (1) general orientations toward the news media as an (authoritative) source of information, (b) integration of various alternative sources of information, and (c) expressions of agreement and disagreement. Using the three dimensions as analytic points of departure, the inductive empirical analysis focused on mapping the different ways in which belief validation occurs during conversations. In the results section, all quotes are anonymized. When required for illustrating conversational dynamics, a number identifier (#1, #2, etc.) is used to differentiate between speakers.

Results

Figure 2 presents descriptive statistics on how news coverage of crime and crime perceptions developed across the six panel waves between early 2020 and late 2022. With respect to news coverage, not only did the number of articles focusing on violent crimes grow over time, but the tone of coverage became increasingly negative—both in terms of the presence of a negative tone in the news and whether negativity dominate the stories. Seen from the start, the share of stories with a mainly negative tone increases from 16% (W1) to 42% (W6). Crime beliefs became increasingly negative overall as well. The perception that Sweden has very large problems with violent crimes is shared by 44% in Wave 1, up to 60% in Wave 6. The belief that more violent crimes are committed in Sweden than in most other EU countries, goes from 14% to 26% during the same period. A significantly smaller share believes that the number of crimes has decreased in Sweden in the past years, but the level remains fairly stable over time.

If reinforcement is conceptualized as a (long-term) process belief intensification over time, there is first indication for aggregate belief reinforcement as well. Using the averaged crime belief index of these three items, ranging between 1 and 7, the public displays slight negativity in Wave 1 (M = 4.59; SD = 1.42), which then intensifies until Wave 6 (M = 5.09; SD = 1.43), reaching a top in Wave 4 (M = 5.22; SD = 1.39). It can also be noted that 26% of the sample talks about crime with family and friends at least “several times a week” in Wave 1. This share increases to 32% in Wave 6, after also reaching a top in Wave 4 (37%). In addition, these conversations are overwhelming negative already at the outset. In Wave 1 69% talk primarily about negative aspects of crime—a number that increases to 77% in Wave 6.

The long-term process of belief maintenance and reinforcement

Table 1 presents findings from the unconditional LCM-SR speaking to belief reinforcement and reciprocal influences over time. A few things are particularly worth noting. First, the model fits the data well in terms of fit indices (RMSEA = 0.054; CFI = 0.969). Second, the between-person part of the model shows that there is indeed evidence for long-term belief reinforcement as the latent slope is positive and statistically significant (M = 0.11, p < .001). This means that, overall, pessimistic beliefs intensify over time. In contrast, a non-significant slope would have reflected overall belief stability (maintenance). Furthermore, the latent slope of talk also displays significant growth over time (M = 0.02, p < .001), reflecting a parallel increase in negative crime talk over the same time period. The covariances between the two latent intercepts (COV = 1.06, p < .001) and the two latent slopes (COV = 0.01, p < .001) for beliefs and talk further suggest that these factors are positively related, both with respect to levels and developments over time: respondents who generally talk more negatively about crime also view current conditions more pessimistically and increases in talk are related to increases in negative beliefs over time. Third, the within-person part of the model shows that the only significant intra-individual effect is the instantaneous impact of talk on beliefs (b = 0.07, p < .001). Thus, the immediate impact of more negative talk about crime with family and friends, is more pessimistic beliefs about crime developments in society. Overall, these findings suggest a clear process of long-term belief reinforcement in which interpersonal crime talk plays a role, both at the between- and within-person level.

To both assess the sensitivity of these findings and explore the role played by individual-level news exposure, several alternative model specifications are presented in the Supplementary Appendix. First, a conditional LCM-SR with several between-person controls (gender, age, education, political interest, and ideology), as well as inter-individual differences in traditional news media use, was tested (Supplementary Appendix Table A1). Second, a tri-variate RI-CLPM was tested focusing on the dynamic within-person reciprocal relationship between traditional news media use, crime talk and crime beliefs over time (Supplementary Appendix Table A2). Apart from replicating the findings presented above, both these two alternative models show that while traditional news media use is unrelated to crime beliefs, news exposure is positively related to crime talk. The tri-variate RI-CLPM further reveals positive reciprocal relations between traditional news exposure and crime talk, above and beyond a distinct positive effect of crime talk on beliefs. Third, a LCM-SR using an alternative operationalization of interpersonal talk—focusing on discussion frequency only, not weighted by tone—was tested, yielding very similar findings as the main model presented here (Supplementary Appendix Table A3). Finally, several longitudinal mediation models were estimated to assess the indirect effect of traditional news media use on crime beliefs, via interpersonal talk (Supplementary Appendix Table A4). In sum, these findings suggest that the primary role of news exposure is not to directly influence crime beliefs, but rather to stimulate crime talk in a reciprocal manner and thus indirectly influence crime beliefs.

The social-conversational practices of message and belief validation

While the panel analyses lend support for the filtering hypothesis and the role of interpersonal communication—that talking about violent crimes in ways consistent with dominant patterns of news coverage contributed to belief reinforcement—these data reveal little about what is actually going on in such conversations and, more specifically, how messages and beliefs are validated when citizens talk to each other. Based on an inductive close reading of the discussions guided by the three theoretically derived dimensions of social communication, several social belief validation practices were identified. As already demonstrated by the quantitative, representative, survey data, conversations about crime were heavily dominated by a focus on negative conditions—both reflecting and validating patterns of news coverage. Also, the group discussions are permeated by negativity. But the way that media messages and beliefs about crime are validated varies. Figure 3 summarizes the main findings from the qualitative analysis by suggesting distinct concepts, describing their main functions in the conversations, and briefly illustrating their different manifestations in the data. While our main focus is social belief and message validation practices, the data also reveal several instances of their counterparts: social rejection mechanisms, which are also outlined in Figure 3. Importantly, each of these validation practices reflect more specific dimensions of the filtering function as broadly captured by the survey measures presented above.

Level 1: referential validations (and critical rejections)

The most straight-forward way in which the negative worldview of violent crimes portrayed by the news media is socially confirmed us through referential validation. A referential orientation entails an uncritical acceptance and expression of mediated accounts of reality, while a critical orientation cast doubt over, or explicitly question, crime reporting in the news media. Referential validation can occur either through subtle aligning, whereby participants talk about crime in ways that reflect the dominant patterns of news coverage, or through explicit referencing, as participants more directly refer to media reports.

Although these are distinct practices at the conceptual level, real-world conversations are dynamic and fluid, moving back-and-forth between these orientations. One example comes from an elderly group (aged 60–75 years). While a referential orientation dominates the discussion, validating an overall negative account of reality, a critical orientation is sometimes suddenly activated. For instance, following several referential exchanges about crime developments, referring to media reports, one respondent expressed an indication of doubt, but without challenging the main lesson learned from the media:

At the same time, between the lines you learn that there are no more death shootings now than ten years ago. what are you supposed to believe? That’s what I wonder. But reports in the media make you believe that death shootings and explosions are increasing… but whether it is true, I don’t know… but this is the main concern of course.

The conversation then quickly gravitates back to a referential orientation, with explicit referential validation of negative crime developments:

That’s how it is… and it’s not only that you see it on news programs or read about it in newspapers, but they even have special programs about crime and crime bulletins and whatever it is called on TV aiming to reflect this…

While referential orientations permeate the discussions, critical orientations are not uncommon—and in the current case, critical rejections primarily serve the function of questioning the negative picture of crime in the media. Manifestations include both rather non-elaborated “automatic” expressions of doubt about the media as well as more reflective elaborations of news coverage as being biased. Even within the most critical conversations, however, several participants admit that the negative picture of crime portrayed by the media is very difficult to escape from or remain immune to, “despite that there are positive things happening”—indicating the power of dominant narratives in the media.

Level 2: resonance validation (and dissonance rejection)

A second way in which beliefs are validated during conversations relates to how participants connect media coverage to other forms of knowledge and experiences. This can be done in ways that either amplifies the negative portrayal of crime provided by the media (resonance) or in ways that challenge the news medias’ account of reality (dissonance). Compared to the referential-critical orientation dimension, this conversational practice focuses on the integration of additional information into the discussion. Following the data analyses, I propose resonance validation (and dissonance rejection) as specific conversational practices that amplifies (or challenges) the dominant media narrative. While resonance validation entails providing additional “evidence” supporting mediated accounts of reality, dissonance rejection entails the opposite—providing additional “evidence” questioning media coverage. Again, this can be done in various ways, as evident in the data, including drawing from personal stories and experiences of friends, relatives or acquaintances, to knowledge acquired from other sources of information than the news media.

Experiences supporting a negative account of reality enter conversations in different ways. A first observation is that although experiences of crime are not uncommon in the groups, they oftentimes emerge somewhat disconnected from broader societal conditions—as personal stories or anecdotes seemingly unrelated to events and trends reported by the media. Other times, however, participants explicitly relate experiences to broader trends. One illustration of how experiences of significant others enter the discussion in ways that resonate with a negative account of crime developments, comes very early in one of the groups.

I think a little bit about generations before us… who have told… my parents who have told about the difference from when they grew up until today… when we are growing up… then I very often hear that in terms of crime it used to better before… and what NN [previous speaker] mentioned that it has escalated… it becomes more and more… it increases all the time… in the past you didn’t have to lock the door… kind of… and could feel safe… but that’s not possible anymore.

Another source of evidence is based on a different point of reference than personal experiences: “statistics”. On the one hand, “statistics” is primarily used to either challenge or provide more nuance to the negative worldview provided by the media (i.e., dissonance rejection). On the other hand, references to statistics mainly serve as a secondary, or complementary, piece of information to counter-argue what is perceived as a strong, omnipresent, negativity bias in the media. As such, perceived inconsistencies between these sources of information trigger significant uncertainty regarding what is actually “true”—manifested both at the individual and group level. Asked whether developments are going in the right or wrong direction, all participants in one of the groups say “wrong”—illustrating both the immediate accessibility of negativity beliefs (almost triggered automatically) and the complete group consensus. Immediately following this disclosure, one participant expresses what appears to reflect a state of cognitive dissonance related to “statistics”:

But at the same time I believe there are statistics pointing towards that at least some crimes are moving in the right direction… if one may say… that they are going down… but my view is that even though statistics tell differently… it is still getting worse, kind of…

The exact source of such “statistics” are seldom clarified—potentially originating from more close readings of news reports among engaged citizens—but these numbers are still referred to as evidence independent from the media.

Level 3: social validation (and social rejection)

While the first two dimensions relate to the content of statements made during conversations, the third level proposed here focuses on social reactions to such statements, as implicit or explicit manifestations of agreement and disagreement. Thus, conversational practices of social validation (and social rejection) are inherently “social” in nature. At the conceptual level, social validation refers to group interactions validating participants’ statements, while social rejection does the opposite—questioning participants’ statements. While these are generic social practices applicable to any conversational situation, they emerge in talk about crime developments and whether current conditions are negative or positive as well. As such, practices of social validation (and rejection) are visible in several ways.

One common way in which agreement is expressed is through explicit statements of support. This happens continuously as participants respond to and refer to previous speakers. Examples from the group discussions include statements like “I agree with everybody else on this point”, “I agree with NN regarding integration.”, “me and NN have talked about this long ago… and I agree with her…”. These are simple but distinct forms of social validation that permeate all discussions of crime. In contrast, explicit statements of disagreement are completely absent (e.g., “I do not agree with NN”) in the data. This does not mean that other forms of disagreement do not surface.

Another way by which beliefs are validated occurs through “associative confirmation”—creating a shared form of mutual storytelling and understanding among participants. This also happens frequently in the discussions. Talking again about negative crime developments, the following conversation sequence illustrates this well.

[#4]“I would also [referring back to previous claims] say that that it is going in the wrong direction, as we said in the beginning and then I think that it depends a lot on… what NN [earlier speaker] referred to earlier… that many civilians are affected and this was not the case in the past… is my impression…”

[#2] “When is ‘in the past’, if you compare to the past?”

[#4] “Yes, but kind of what you say… that you… if you talk to your parents or something… but also… then I don’t know if the media have changed their way of writing about it either… when I was younger in my teens… then there were not as much shootings affecting others… then it was within the criminal gangs… so it… not even that long ago I think… but 10-20 years…”

[#2] “Yeah, that was kind of what I was thinking as well before…”

[#4] “That recently it more and more affects… others than the criminal gangs sort of…”

This reflects an instance of associative confirmation validating a negative account of reality. Disagreement—and social rejection—is not absent in the discussions, however, although significantly less prominent. Practices of social rejection can be both “soft”, subtly questioning others’ statements through expressions of doubt rather than outright challenge, as well as more direct, explicit, and confrontational. Such direct and explicit questioning is not very frequent, but does occur. The social dynamic of such sequences of disagreement often gravitates towards areas of agreement, reflecting attempts to avoid conflict and find common ground.

***

While the quantitative (representative) study shows that citizens talk about crime in ways that reflect the dominant patterns of news coverage, validating and reinforcing pessimistic crime beliefs over time, the group discussions reveal the more distinct ways in which such filtering and validation occurs in social communication. Apart from talking about crime developments in ways that align media coverage (referential validation), mediated accounts of reality are supported through other forms of “evidence” (resonance validation), and through interactive confirmation from conversation partners (social validation). These practices and mechanisms are summarized in Figure 3, along with various rejection mechanisms also present in the data.

Conclusion and discussion

This study aimed at making a distinct contribution to research on media effects, focusing on more long-term effect dynamics—described by cultivation theory but seldom addressed empirically. Leveraging a unique mixed-method data collection combining quantitative and qualitative approaches, this study analyzed belief maintenance and reinforcement as outcomes of media coverage and interpersonal communication. Following cultivation research by focusing on news reporting and beliefs about crime developments in Sweden, the analyses looked both into the (a) long-term processes and the (b) social validation mechanisms behind these effect dynamics. Let us briefly summarize the main findings and discuss their implications.

With respect to the long-term process, the results show that increasingly negative news coverage of crime coincided with increasingly negative beliefs about crime developments among citizens over a period of almost three years (2020–2022). During the same time, citizens also discussed violent crimes more frequently and pessimistically in their everyday lives—altogether reflecting parallel aggregate trends in news coverage, interpersonal talk, and beliefs. These factors were also related at the individual level. Using LCM-SRs to capture both immediate effects and long-term dynamics, the findings revealed a clear pattern of belief reinforcement over time associated with increases in negative interpersonal talk, both at the within- and between-person level. In addition, individual news exposure was significantly related to interpersonal crime talk, but not to beliefs, suggesting an indirect effect dynamic operating through interpersonal talk. With respect to the social validation mechanisms underpinning those reinforcement effects, analyses of group discussions identified several ways by which crime beliefs and media messages are validated during conversations—reflecting specific filtering functions of interpersonal communication. Along three dimensions, I proposed referential validation (and critical rejection), resonance validation (and resonance rejection), as well as social validation (and social rejection), as distinct mechanisms of social communication reinforcing negative news about crime in the media. These findings have implications both for research on long-term media effects as well as the filtering function of interpersonal communication.

First, the study contributes to research on long-term effect dynamics. Studies of media effects have been dominated by a focus on immediate, short-term, change as the main outcome of interest, even though several theories assume more long-term influences. The classic case is cultivation theory, according to which repeated exposure to a continuous flow of dominant media messages cultivates (shapes and reinforces) beliefs and worldviews among citizens (Potter, 2014; Shrum, 2017). But also classics like agenda-setting (McCombs & Valenzuela, 2020) and framing (Baden & Lecheler, 2012; Shehata et al., 2024; Tewksbury & Scheufele, 2009) refer to long-term influences. By using a longitudinal design spanning over almost three years, and statistical methods directly speaking to the relevant conceptual properties outlined, the study documents such long-term processes empirically, at both the within- and between-person level. As such, the findings support theories claiming that beliefs (and attitudes) are reinforced in response to a consonant increase in news coverage across the most important media outlets. When news coverage of crime intensifies and becomes more negative over time, negative crime perceptions also intensify. This happened during a time period when the number of confirmed cases of lethal violence did not increase in society, although the character of such crimes had changed significantly—with growing numbers of shootings and use of firearms—as well as a more long-term increase in lethal violence (Brå, 2023a, 2023b; Kunskapsverket, 2023), which gained significant attention in the media. This is also evidence that broad patterns of news coverage still matter for public opinion formation—even when actual crime trends are not as clear-cut—despite growing audience fragmentation and content personalization (Bennett & Iyengar, 2008; Djerf-Pierre & Shehata, 2017; Geiß, 2022). In that sense, some of the core tenets of cultivation theory in terms of shaping crime perceptions may remain relevant in contemporary societies, given certain contextual conditions.

Second, the study highlights the role of interpersonal communication for these long-term effect dynamics. In line with theories of social communication and the filtering hypothesis (Schmitt-Beck & Lup, 2013; Song & Boomgaarden, 2017), the influence of news coverage appears particularly strong when media messages are validated in conversations with friends and family. Along with media messages, beliefs and attitudes are continuously negotiated within social networks as part of a people’s everyday lives (Katz & Lazarsfeld, 1955; Neiheisel & Niebler, 2015). This study linked these social validation mechanisms to long-term processes of belief reinforcement, providing empirical evidence for the conditional nature of these broad effect dynamics (Valkenburg & Peter, 2013). Moving beyond quantitative measures of interpersonal talk, qualitative analyses of crime conversations unpacked the more specific ways in which the filtering function of social communication operates: through (a) referential validation, (b) resonance validation, and (c) social validation. As such, the study has not only provided unique empirical evidence supporting the filtering function of interpersonal communication for long-term belief reinforcement effects, but also documented how such belief validation (and rejection) practices occur during conversations (Conover & Miller, 2018; Gamson, 1992) – making a distinct contribution to research on the filtering hypothesis as well (Neiheisel & Niebler, 2015; Schmitt-Beck, 2003; Schmitt-Beck & Lup, 2013). Thus, the instantaneous and long-term filtering effect on belief reinforcement documented in the quantitative part of the study, is not only about citizens repeating a negative crime message in everyday social communication (referential validation), but also about integrating personal experiences and other sources of information (resonance validation) as well as confirming each other interactively (social validation). Although these practices have partly been documented in previous research (Conover & Searing, 2005; Gamson, 1992) the current study links them directly to the filtering function of media effects, focusing on the specific case of crime news.

Third, in several ways, the study focused on conditions, processes and outcomes outlined by cultivation theory: long-term effect dynamics, consonant media messages across news outlets, as well as belief maintenance and reinforcement of crime beliefs (Morgan et al., 2012; Shrum, 2017). This “one-sided”, consonant, news environment represents a particular but not uncommon context for media effects. Although the concept of resonance has played a significant role in cultivation research, referring primarily to a “double dosage” of media messages and personal life experiences, interpersonal communication has not received much attention on its own in cultivation research. As such, the integration of interpersonal communication as a validation mechanism can be seen as an extension of cultivation research in relation to the concept of resonance. Social communication is a crucial mechanism through which various life experiences—personal as well as those of family, friends, and acquaintances—are shared as part of everyday conversations. Thus, this study illustrates the potential of extending the concept of resonance to address the filtering function of social networks in studies of long-term cultivation effects (Gamson, 1992; Shehata et al., 2021).

While the filtering function of interpersonal communication for belief reinforcement is potentially crucial in this regard, it also raises the question of how these mechanisms operate in a social media context, where much mediated interpersonal communication takes place—something not addressed in this study. Both as a source of news and a platform for commentary and discussion, social media may well influence exposure to crime news as well as how beliefs are validated in social interactions. Without downplaying social networking sites, however, traditional news media remain, by far, the most important sources of news for the vast majority of Swedish citizens (Anderson, 2023; Nygren & Widholm, 2023). With respect to the filtering function of social networks, however, engagement on social media may both resemble and deviate from offline interpersonal communication. On the one hand, all three filtering mechanisms identified here—referential validation, resonance validation, and social validation—are applicable also to interactions on social networking platforms where users consume and try to make sense of news. For an issue such as crime, potentially highly salient on these platforms, this is even likely. On the other hand, research suggests that individuals primarily discuss political and societal matters with people in their close-tie networks: thus being the primary contexts where beliefs are socially validated and negotiated (Ekström, 2016; Mutz, 2006). Nevertheless, how the filtering function operates on social media is a promising area for future research, not least given the relevance of personalized reinforcing spirals on these platforms (Slater, 2015; Song & Boomgaarden, 2017).

While this study provided unique evidence concerning the role of interpersonal communication for long-term effect dynamics, building upon quantitative and qualitative data analyses, future studies would gain from a stronger longitudinal match of the qualitative and quantitative approaches. The current study analyzed group discussion data from a single point in time, identifying various validation mechanisms relating to the filtering function of social communication, but whether such practices change over time, or vary depending on changes in media coverage, could not be determined here. Also, additional group discussion data, would provide more opportunities to study how validation mechanism vary not only over time, but also between different groups of citizens.

Supplementary material

Supplementary material is available online at Human Communication Research online.

Data availability

The data underlying this article cannot be shared publicly due to data restriction policies.

Funding

This work was supported by the European Research Council (ERC) under the European Union’s Horizon 2020 under grant agreement no. 804662. Any opinions, findings, and conclusions or recommendations expressed in this article are those of the author(s) and do not necessarily reflect the views of the ERC.

Conflicts of interest

None declared.

References